Some researchers spend all their professional lives at research facilities because they can spend all their time innovating without being bound by business constraints, such as time to market. According to Krishna Dubba, CTO and co-founder at event sponsorship platform provider CoVent, while both disciplines provide continuous learning opportunities, the intersection of two mindsets can serve businesses and their CIOs well.
Dubba grew into the role of a researcher after earning a series of IT-centric degrees, starting with a bachelor of computer science degree at Jawaharlal Nehru Technological University. Near the end of the program, he took the only AI courses available in 2004.
“At that time, I realized I needed to study AI more because it was so fascinating, and there was only one university offering a master’s degree in AI in India at that time. It was the University of Hyderabad, which is very research oriented,” says Dubba. “I did a lot of research at the end of the course, trying to find computer viruses using AI techniques.”
Next, he went to work for a hedge fund company working on algorithm development before deciding to go back to school at the University of Leeds to pursue a Ph.D. on a grant from The European Research Commission that covered all costs.
“The European Research Commission is a group of countries that work collectively on research. I got the opportunity to go to different countries and work with different universities,” says Dubba. “I was trying to analyze what is happening in videos using computer vision. It’s easy for humans to understand what is going on, but it’s very difficult for a machine.”
His Ph.D. work focused on “cognitive vision” that allows a machine to recognize objects and comprehend what’s happening in a video. One project involved an airplane at an airport. When a flight lands, a lot of activity ensues on the ground, at the gate, on the plane, and more. Using cognitive vision, the airline was able to identify process inefficiencies that could be used to lower costs and improve safety.
However, as a post-doctorate, Dubba chose a more difficult challenge: Robotic vision.
“During my Ph.D. we used fixed cameras to record video, but with a robot, you need ‘egocentric vision’ because the cameras move with the robot. So, as a post-doc, I could see that everything becomes a more complex understanding of what’s going on in a video,” says Dubba.
From Research Facilities to Startups
Many different industries can use cognitive and egocentric vision for their benefit, so Dubba went to work as a principal researcher at Nokia Tech’s Advanced Research Lab.
“At the time, Nokia had an eight-camera device, called OZO, that looked like the head of a duck and cost around $50,000. Its purpose was to capture 360-degree views. If you wore a headset, you could experience the video in 3D, meaning you could look up, down, or in any direction to explore it,” says Dubba. “It’s called, ‘presence capture,’ and there were a lot of problems with it because you have to stitch the video from eight cameras together.”
Next, he worked for Nokia Bell Labs solving problems in deep learning. Dubba worked in the Social Dynamics group that is charged with quantifying the unquantifiable.
“[Research facilities like Bell Labs] don’t ask, ‘How fast can we build or how much money can we make?’ They want to understand how it will change humanity, so most projects have a 10-year lifespan,” says Dubba. “We were trying to measure things like the emotion or health of a city, which is a very challenging problem, because it is hard to define and hard to measure, so you must use proxies. We used social media feeds as proxies. For example, a map app can tell you the fastest way to get to a destination easily because it is easy to define and measure, but it can’t tell you what your ‘happy path’ would be as it is vague, subjective, and hard to measure.”
Next, he went to work for automation company Blue Prism, first as a senior research scientist and later as a staff research scientist. At the time, the company was building robotics process automation solutions so organizations could automate business processes. When Dubba joined, the company had recently created an AI lab in London. His job was to set the strategy and recruit the researchers.
“It was fascinating — completely different from what I did before. As a researcher, I never had to worry about the business impact or justify the business case, so I learned how to determine the value of a research project from a commercial point of view,” says Dubba. “We also wrote a lot of patents, four of which were granted by the time I left. We also built a product called, ‘Capture,’ so fast, I realized I was working in an entrepreneurial environment. So, I started thinking that I could found a startup doing the same things.”
The first two companies Dubba co-founded were AI-powered life balance app provider Jeevi AI and enterprise-grade GenAI solution provider A2O. Both companies failed for a common reason: A lack of domain expertise.
“Jeevi AI had 25,000 customers, so we were monetizing the product, but we couldn’t make enough to sustain the business. At A2O, we built a chatbot that allowed users to ask questions on unstructured data and documents using LLMs. We also built a product called, ‘Insights’ that used the structured data users fed it so the users could ask data science queries using natural language,” says Dubba. “But we realized that we were not domain experts, we were all technologists.”
So, for CoVent, Dubba decided to co-found the company with a sales and marketing expert who happened to have the expertise it would take to shape a product that could sell. CoVent is an event sponsorship platform provider that helps event producers and sponsors to find each other through the platform versus using Google Search. There are also a lot of holes in the data because different event producers present their offerings in different formats. CoVent normalizes the data, so it’s easier to understand individual events in greater detail, faster than search.
The company is currently funded by Founders Factory, a startup accelerator and Founders Factory’s new Pico Venture Studio.
“In a commercial setting, you have to quickly validate your assumptions, which is where the research mindset comes in. You start with a lot of assumptions like the user might pay X for a feature, and the only way you can validate it is to get it out there and get feedback. That’s why Mark Zuckerberg says, ‘fail fast,’” says Dubba. “The whole thing is about iteration and understanding what customers want and don’t want and why.”
Continuous learning has also helped him stay current with technology as a hands-on CTO and co-founder. For example, at Jeevi AI, he was the architect and principal backend Python developer. AT A2O, he was the system architect and principal backend ML engineer developing Generative AI solutions for deeper business insights using LLMs.
His journey has also taught him how to lead effectively.
“I always try to take an empathetic route, because that’s what I would want if I were the employee. There are a lot of reasons why a person may feel a certain way at any given moment. And people will respect you more as a leader and work harder for you when you’re willing to give them a second chance to succeed,” says Dubba. “When I was a researcher, I would work on part of a problem as a group member, but as a CTO, what I do now directly impacts peoples’ lives. I think being empathetic helps create a healthier company environment. And for me, personally, my research background helps me balance the desire to innovate with the need to validate the commercial viability of a solution.”
As a researcher, he also learned the value of resilience and iteration.
“Experiments often fail, so it’s important that CIOs see failure as feedback so you can improve on the idea,” says Dubba. “Academic research thrives on intersections of disciplines. Adopting this mindset can encourage CIOs to break down silos, mix methods and connect the dots across departments. For example, the best AI ideas came from biology and physics.”